[Test]
public void AllMappingsAreOk()
{
	var session = GetCurrentNHibernateSession();
	var errors = new List<string>();

	foreach (var entry in session.SessionFactory.GetAllClassMetadata())
	{
		var type = entry.Key;
		var data = entry.Value;

		try
		{
			session.CreateCriteria(type).SetMaxResults(0).List();
		}
		catch (Exception e)
		{
			errors.Add(string.Format("Error getting type {0}: {1}", type, e.Message));
		}

		foreach (var collection in data.PropertyNames.Where(x => data.GetPropertyType(x).IsCollectionType))
		{
			try
			{
				session.CreateCriteria(type).CreateCriteria(collection).SetMaxResults(0).List();
			}
			catch (Exception e)
			{
				errors.Add(string.Format("Error getting collection {0}.{1}: {2}",
										 type,
										 collection,
										 e.Message));
			}
		}
	}
	Assert.True(errors.Count == 0, string.Join(Environment.NewLine, errors));
}

This piece of code was really helpful in the past and is following me in every project involving NHibernate. It seemed useful to share

• Circular references support
• Support for .NET 3.5 and Mono 2.0

Thanks to Yallie for the work on supporting more targets than 4.0!

Here is the release: http://memodb.codeplex.com/releases/view/70652

SELECT * FROM Activities a Where a.Community in ('Community 1', 'Community 2', 'Community 3')

…with Linq.

In the objects world this is managed using the Contains method of IList. The perfect Linq for the previous query is:

var communities = new []{ "Community 1", "Community 2", "Community 3" };
var result = activities.Where(x => communities.Contains(x.Community));

The problem with this is no all Linq providers understands it. And SPLinq (Linq for SharePoint) is one of those who doesn’t.

The other way to handle this kinds of queries is, in SQL:

SELECT * FROM Activities a Where a.Community = 'Community 1' or a.Community = 'Community 2' or a.Community =  'Community 3'

But how to create the Linq query dynamically? Well… All these months working with Magiq gived me the ability of compile Expression trees in my head

This is the code needed:

private static IQueryable<T> Where<T>(this IQueryable<T> self, Expression<Func<T,object>> property, IEnumerable<string> @in)
{
	var propertyName = ((MemberExpression)property.Body).Member.Name;

	Expression expression = null;

	var x = Expression.Parameter(typeof (ActivityItem), "x");
	foreach (var community in @in)
	{
		var propertyExpression = Expression.Property(x, propertyName);
		var equal = Expression.Equal(propertyExpression, Expression.Constant(community));
		expression = expression == null ? equal : Expression.Or(expression, equal);
	}

	if (expression == null)
		return list;

	var lamda = Expression.Lambda<Func<T, bool>>(expression, x);
	return self.Where(lamda);
}

var communities = new []{ "Community 1", "Community 2", "Community 3" };
var result = activities.Where(x => x.Community, @in: communities);

Linq rules!!

First I tried to work with the MetadataWorkspace and I gave up, ’cause its really impossible to get the mapping information. Luckily, I found Linq to Edmx project who does exactly what I needed.

Then I had some troubles working with the Queryables generated by the ObjectSets. Basically, the expression generated calls methods that doesn’t work and are there only to give information in the expression. So the expression is not something you can compile and execute. As big part of Magiq is to manipulate expression trees, that was a really big problem. I needed to figure out how to change the source queryable (I mean, the queryable where all the methods are applied to) to a simple ObjectSet. Besides it works now, I’m still working to get to a nicer way (you don’t want to know what I have done to make it work…).

Since I changed the insert API from having the new item expression (the new{…}) outside to have it as a select, I had to break the query in two, the select and the “rest”. Ok, that gave me more headaches .

But the biggest problem was when Entity Framework started to change the order of the columns in the select (and add new columns). Let me explain how Linq to Sql works:

If you do this:

items.Where( ... ).Select( x => new { A = x.Prop1, B = x.Prop2, C = x.Prop2 /*again*/ });

Linq to sql transate it to:

SELECT prop1, prop2, prop2 FROM items WHERE ...

But Entity Framework removes the repeated columns and maybe reorders them, giving you something like:

SELECT prop2, prop1 FROM items WHERE ...

You don’t have any idea of the reflection tricks I had to make to get those property-select columns mapping.

But well, those are not bad news at all. I mean I could work around them and I can say Magiq now supports Entity Framework (with a few restrictions). Those are:

• Only Table per class hierarchy inheritance is supported. I’m currently working on the Table per concrete class and Table per type strategies
• Only mapped associations are supported. This means associations based on foreign keys are not (yet).
• Only mapped POCOs are supported. This means Self Tracking entities and Code First are not (yet).

Keep in touch for the updates

For those reading this for the first time (?), Magiq is a framework that let you do mass operation in a Linq fashion.

So… as Magiq will be supporting Entity Framework soon, it made sense to change the API to make it support Entity Framework limitations. Also, I found a better Insert API (the one that always made me doubt!). The changes are:

Query extension method changes
The new method looks like this:

country.Query(x => x.Cities);

It will return an IQueryable that will let you do all the Magiq stuff.

No more IEnumerable support
This is a hard one. You can only do Magiq stuff between IQueryables. If you have an IEnumerable collection, use the Query extension method of it parent entity.

New Insert API

destination.Insert( source.Where( condition ).Select( new Item{ ... } ));

Cool, isn’t it?

Entities extension methods
Because we no longer support extension methods over collections but over the entity, there are some extension methods for improving the code readavility.

Delete

entity.Delete(x => x.Collection.Where( condition ));
//Instead of
entity.Query(x => x.Collection).Where( condition ).Delete();

Update

entity.Update(x => x.Collection.Where( condition ).Set(x => x.Prop, value));
//Instead of
entity.Query(x => x.Collection).Where( condition ).Set(x => x.Prop, value).Update();

Query

entity.Query(x => x.Collection.Where( condition ));
//Instead of
entity.Query(x => x.Collection).Where( condition );

Insert

entity.InsertInto( x => x.Collection, source.Where( condition ).Select( new Item{ ... } ));
//Instead of
entity.Query(x => x.Collection)
      .Insert( source.Where( condition ).Select( new Item{ ... } ));

WeakDictionary to the rescue!

This dictionary, basically, maintains each value in a class that could reference it as a weakreference if it wasn’t accessed often and if the amount of items is bigger than a threshold defined:

var d = new WeakDictionary<int,SomeClass>(3);
d.Add(1, new SomeClass());
d.Add(2, new SomeClass());
d.Add(3, new SomeClass());

//We use the values...
var bla = d[1];
var bla1 = d[1];
var bla2 = d[1];
var bla3 = d[2];
var bla4 = d[2];
var bla5 = d[3];

//When we try to add an item and we are passing the threshold, it will weak the less used reference
d.Add(4, new SomeClass());

d[3] value is now maintained with a weak reference. If it’s used before it’s garbage-collected, it will be alive again and d[2] will be the “weak one”.

You can download the code here.

I started to play with Entity Framework 4 and realized some stuff is not possible the way it was designed. It has to do with the way EF4 manages Lazy collections.
In Linq-to-sql and NHibernate, the collection itself has the property of being lazy, materializing all the objects when you iterate it. For example, in Linq-to-sql collections are EntitySets. Because of this, you can do this:

parent.Children.Delete(x => x.SomeCondition);

This is because the Children collection contains all the information needed:

• The parent object instance, for getting the Id
• The collection name

Sadly, EntityFramework handles this feature in a different way: Is the parent object the lazy one and the collection is populated when you access the collection propery and not when you iterate it. Because of that, in the moment you do parent.Children you have executed the query in the database.
This lead us to have a different API, that avoid accessing the collection property. What I thought is:

parent.Delete(p => p.Children.Where(x => x.SomeCondition));

It’s not so bad, I must Say, and it also let us do things like

parent.Delete(p => p.SomeObjectInTheMiddle.Children.Where(x => x.SomeCondition));

without executing the query for getting the SomeObjectInTheMiddle instance.

Also, I give up with Magiq-to-NHibernate, at least for a while. NHibernate has huge features, it let you map your classes in thousand of different ways, and that is awesome. But also it makes it really complicated to work with. And since NHibernate already has mass operation support, it makes sense to stop worring. At least, as I said before, for a while.

So, I will be working on changing the Query API and only for Linq-to-Sql and Entity Framework. Everyone is welcome to contribute implementations for other ORMs.

Next step: Changing the API

• Inserts
• Updates
• Delete
• Insert references and collections by cascade
• Delete collection children by cascade (aka cascade-all-delete-orphans)
• EntityRef references
• Id auto generation (With defaults for ints, long and guid)

You can download it from here.

To see other posts of MemoDb:

• Introduction to MemoDb
• MemoDb: Collections support
• MemoDb: Auto assign id support
• MemoDb: Cascades and Collections support

To follow the MemoDb posts saga:

• Introduction to MemoDb
• MemoDb: Collections support
• MemoDb: Auto assign id support

Now, MemoDb supports:

• Delete
• Collections
• EntityRef (for LinqToSql support)
• Cascade inserts
• Cascade delete on collections children (aka all-delete-orphans)

Delete
Pretty simple, it let you delete objects:

using (var s = memo.CreateSession())
{
	var city= s.GetById<City>(1);
	s.Delete(city);
	s.Flush();
}

Cascade Inserts
You can configure MemoDb to insert new references by default:

var memo = new Memo(insertByCascadeByDefault: true)

Or you can specify the behaviour just for one relationship:

 var memo = new Memo()
           .Map<City>(c => c.CasacadeInsert(x => x.Country))
           .Map<Country>();

Cascade Delete Orphans
You can configure MemoDb to delete collections children by cascade when the parent object is deleted by default:

var memo = new Memo(deleteOrphansByCascadeByDefault: true)

Or you can specify the behaviour just for one collection:

 var memo = new Memo()
           .Map<City>()
           .Map<Country>(c => c.CasacadeDeleteOrphans(x => x.Cities));

• Introduction to MemoDb
• MemoDb: Collections support

Today I will talk about the generation of ids for new entities. As MemoDb stands for mocking real linq orms and, generally, the underlying databases auto assign the ids.

var memo = new Memo()
     .Map<MyEntity>(x=>x.Id(x => x.Id, autoAssign: true));

By now, MemoDb supports:

• Int32: increments the last id.
• Int64: increments the last id.
• Guid: Generates a new Guid.

If you need another strategy or to map another id type, you can specify the strategy as a delegate:

var memo = new Memo()
     .Map<MyEntity>(x=>x.Id(x => x.Id, last => last + 10));

In the example above, the generation is made adding 10 to the last id assigned.
By default, MemoDb will assign the ids when the object is inserted. You can override this behavior by configuration:

var memo = new Memo(assignIdMode: AssignIdMode.OnFlush)
     .Map<MyEntity>();